Persistent Mapping of Sensor Data for Medium-Term Autonomy.

Area of Science:

• Robotics
• Autonomous Systems
• Geospatial Data Management

Background:

• Autonomous vehicles require robust mapping capabilities for operation in unmapped environments.
• Efficient storage and retrieval of sensor data are crucial for real-time decision-making.

Purpose of the Study:

• To develop a tool for recording and optimizing costmap data placement on a persistent map for autonomous vehicles.
• To enable autonomous navigation and localization in areas lacking prior mapping data.

Main Methods:

• A novel optimization algorithm considering vehicle odometry, GPS (when available), map consistency, and GPS offset error.
• Development of a system for aggregating and storing processed sensor data into a persistent map.
• Testing on diverse terrains including a test track and a dirt trail, with and without GPS.

Main Results:

• Successful creation of high-fidelity maps in previously unseen environments.
• Demonstrated effectiveness of the mapping tool with and without GPS signal availability.
• Validation of map utility for path planning and GPS-denied localization.

Conclusions:

• The presented tool significantly enhances the autonomy of vehicles in unmapped regions by providing reliable persistent maps.
• The optimization method ensures accurate map data placement, crucial for safe and efficient navigation.
• The system proves valuable for both initial mapping and subsequent GPS-denied operations.